Skip to content
View dineshyadav03's full-sized avatar
😉
hello
😉
hello
7 followers · 33 following

Achievements

Achievement: YOLOAchievement: Quickdraw

Achievements

Achievement: YOLOAchievement: Quickdraw

Block or report dineshyadav03

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
dineshyadav03/README.md

Dinesh Yadav

Forward Deployed Engineer · AI Engineer

I build and ship real-world AI systems — LLMs, agentic workflows, computer vision, and robotics — and work close to the problem: deploying, debugging, and hardening AI products in production. I also contribute fixes upstream to the open-source tools that power that work.

Projects

  • taxcite — citation-grounded RAG over Indian income tax law, implementing all ten standard RAG architectural patterns plus three original ones over a shared retrieval substrate, evaluated against a hand-verified golden set.
  • ai-incident-rag — citation-grounded RAG over real AI/LLM production incident postmortems, with hybrid retrieval, reranking, CI-gated generation quality, and prompt-injection guardrails on free-tier infra.
  • fineai — full-stack platform for fine-tuning Cohere models on custom datasets, with real API integration and a training dashboard.

Merged

  • pipeshub-ai/pipeshub-ai — fixed org signup returning HTTP 500 instead of 400 for validation errors (weak password, duplicate org, invalid domain)
  • zed-industries/zed — fixed a multi-monitor DPI window-placement bug

In review

  • anthropics/claude-agent-sdk-python — fixed ClaudeSDKClient.receive_response() returning before background agent work finishes
  • mlflow/mlflow — fixed a GenAI evaluation-dataset bug where a type mismatch broke Postgres tracking backends
  • griptape-ai/griptape — fixed image/audio MIME-format corruption in the MCP tool integration (approved, pending merge)
  • ggml-org/llama.cpp — fixed completion_tokens/total_tokens undercount in the server when n > 1
  • home-assistant/core — added a missing device-state enum value to the NUT integration
  • astral-sh/ruff — fixed a FURB103 lint false positive using type-inference analysis
  • ollama/ollama — fixed a stray environment variable causing spurious 401 errors in the local LLM server
  • excalidraw/excalidraw — removed obsolete contribution docs

Focus areas

LLMs Agents Computer Vision Robotics AEC Developer Tooling

Pinned Loading

  1. taxcite taxcite Public

    Citation-grounded RAG over Indian income tax law (2025 + 1961 Acts) — exact section-number retrieval, hosted embeddings, evaluated against a hand-verified golden set.

    Python 1

  2. ai-incident-rag ai-incident-rag Public

    Citation-grounded RAG over real AI/LLM production incident postmortems — hybrid retrieval, reranking, CI-gated generation quality, and prompt-injection guardrails, built on free-tier infra.

    Python

  3. claude-agent-sdk-python claude-agent-sdk-python Public

    Forked from anthropics/claude-agent-sdk-python

    Fork of anthropics/claude-agent-sdk-python — my open PR #1224 fixing ClaudeSDKClient.receive_response() returning before background agent work finishes.

    Python

  4. cody cody Public

    AI coding mentor for Claude Code — teaches existing repos and general topics

    TypeScript

  5. postquantumkeyexchange postquantumkeyexchange Public

    Post-quantum key exchange implementation in Python — exploring lattice/code-based cryptography ahead of the classical-cryptography breaking point.

    Python

  6. ggml-org/llama.cpp ggml-org/llama.cpp Public

    LLM inference in C/C++

    C++ 127k 22.8k